Including Neurodivergent Participants in User Research

Short answer: To include neurodivergent participants (for example autistic people, people with ADHD, dyslexia or dyspraxia), make the study predictable and flexible rather than "special": send the questions or topics in advance, explain exactly what will happen, let people choose how they take part (spoken, written, live or in their own time), ask concrete questions one at a time, allow breaks and silence, and ask each participant what helps them. Do not treat communication differences as data-quality problems. Many breakdowns in research sessions come from a mismatch between the researcher's and the participant's communication styles, not from the participant.

This guide explains why neurodivergent people are often missing from research, how to recruit and screen respectfully, how to adapt interviews and usability tests, and how to analyse what you hear without penalising different ways of communicating.

Why this matters

Neurodivergent people are a large share of every user base, whether or not you know it. The US Centers for Disease Control and Prevention (CDC) estimated that about 6.0% of US adults, roughly 15.5 million people, had a current ADHD diagnosis in 2023, and about half of them were first diagnosed as adults. Its autism monitoring network estimated that about 1 in 31 eight-year-olds in the US were identified with autism in 2022, up from 1 in 36 in the previous report. Many more people are undiagnosed, or diagnosed with conditions such as dyslexia that these figures do not include.

Most research sessions are designed around one communication style: live video, open-ended questions asked aloud, quick turn-taking, eye contact and small talk. People who find that format draining or confusing either decline to take part or take part and come across as "poor participants". Either way, their experience of your product goes unheard.

Communication differences are two-way

The autistic scholar Damian Milton named this the "double empathy problem" (Disability & Society, 2012): misunderstandings between autistic and non-autistic people run in both directions, rather than coming from an autistic deficit.

Research supports the idea. In a 2020 study in the journal Autism, Catherine Crompton and colleagues passed a story along "diffusion chains" of eight people (72 participants in total), made up of all autistic, all non-autistic, or mixed groups. Detail was lost significantly faster in the mixed chains, while all-autistic chains did as well as all-non-autistic ones, and rapport was rated lower in mixed chains. The authors concluded that these results "challenge the diagnostic criterion that autistic people lack the skills to interact successfully."

For researchers, the practical lesson is this: when an interview with a neurodivergent participant goes badly, look at the format and the interviewer before you look at the participant.

Recruiting and screening respectfully

1. Decide why you are including neurodivergent participants. If you are researching accessibility or a feature aimed at them, recruit deliberately. If you want a representative sample, make your general study accessible and ask about access needs, rather than asking everyone to disclose a diagnosis.
2. Ask about needs, not labels. "Is there anything that would make taking part easier for you?" is more useful, and less intrusive, than "Do you have a diagnosis?". Many people are self-identified or undiagnosed.
3. Make disclosure optional. If you do ask about neurotype, explain why, make the question optional, and treat it as sensitive personal data.
4. Recruit through trusted channels. Neurodivergent-led organisations and communities can help, but approach them as partners and pay for their time.
5. Write clear invitations. State exactly what the study involves, how long it takes, what format it uses, what you will ask about, how people are paid and what happens to their data. Ambiguity is a common reason people decline.

Adapting interviews

Before the session:

• Share the topics or questions in advance. This helps people who need processing time, and it reduces anxiety about the unknown. It rarely harms the data: you want considered accounts, not surprise.
• Describe the session step by step. Who will be there, whether cameras are on, whether it is recorded, how long it lasts, and what happens at the end.
• Offer a choice of format. Spoken or written, live or asynchronous, camera on or off. Some participants express themselves far better in writing; others find typing slow and prefer to talk.

During the session:

• Ask one concrete question at a time. "Tell me about the last time you set up a new account" works better than "How do you feel about onboarding in general?". Avoid idioms, sarcasm and double questions.
• Be literal and specific. "How long did that take?" rather than "Was that quick-ish?".
• Allow silence and processing time. Do not rush to rephrase. If you do rephrase, say so, so the participant knows the question has not changed.
• Do not interpret body language as data. Lack of eye contact, fidgeting or a flat tone does not mean disengagement or dislike.
• Offer breaks, and accept a shorter session if that is what the participant needs.
• Let people stay on topics they care about for a while. A detailed tangent often contains the most specific insight in the session.

Adapting usability tests

• Reduce sensory load: quiet room, no unnecessary observers, predictable lighting and sound. Remote sessions from the participant's own environment often work better.
• Give written task instructions as well as spoken ones, and let participants re-read them.
• Avoid time pressure unless timing is what you are testing.
• Think-aloud is optional. Talking while doing a task splits attention, which some people find very hard. Retrospective think-aloud (talking through a recording afterwards) or written reflections are alternatives.
• Separate the task from the test: note when a struggle comes from the setup (unfamiliar device, an observer, unclear instructions) rather than your product.

Analysing what you hear

Neurodivergent participants may give very short or very long answers, answer the literal question rather than the intended one, or focus on details others skip. Before treating that as low-quality data:

• Check your question. A literal answer to an ambiguous question is a question-writing problem.
• Value specificity. Detailed, concrete accounts are usually more useful than general opinions.
• Do not filter by "articulateness". Excluding short or unusual answers can quietly remove neurodivergent voices from your findings.
• Look for access barriers in your product, not just preferences. Requirements such as time limits, auto-advancing screens, dense text or forced phone calls are often exactly what neurodivergent participants describe.

Ethics

• Treat neurotype and health information as sensitive data. Collect only what you need.
• Pay people fairly, including for any extra time an accommodation takes.
• Avoid framing neurodivergence as a problem to be fixed in your research questions or reports. Report barriers in the product, not deficits in people.
• Where possible, involve neurodivergent people in designing the study, not only in taking part.

A checklist for your next study

• Invitation states format, length, topics and data use plainly
• Access needs asked about; diagnosis disclosure optional
• Questions or topics shared in advance
• Choice of spoken or written, live or asynchronous participation
• Questions are concrete, one at a time, free of idioms
• Breaks allowed; silence not rushed
• Analysis does not exclude short, literal or unusual answers

How Koji helps

Several of the accommodations above are built into how Koji's AI-moderated interviews work, so you can offer them to everyone rather than on request.

• Spoken or written, and switchable. On the welcome page, participants choose voice or text from the modes you allow, and they can switch between them during the interview without losing the conversation. Someone who starts speaking and finds it tiring can carry on by typing.
• Asynchronous by default. Participants take the interview from a link whenever suits them, in their own environment, with no camera and no scheduled call. There is no fixed session length, so nobody is cut off for taking time to think.
• Predictable and text-supported. The interview explains what an AI interview is before it starts. In voice mode, the interviewer's latest message also appears on screen as text, and rating and choice questions show their options on screen, so participants can read as well as listen. In text mode, Enter sends a message and Shift+Enter adds a new line, so participants can write longer answers comfortably.
• One question at a time, with consistent follow-ups. The AI interviewer asks each question and follow-up in turn, in the same style for every participant, and you set how many follow-ups each question gets (from none to three) in the research brief. That consistency removes some of the interviewer-to-interviewer variation that can make sessions feel unpredictable.
• Structured questions for clear answers. Structured questions, such as rating scales, single or multiple choice, ranking and yes/no, give participants concrete options when an open question would be ambiguous.
• Screening for access needs. Use screening questions to ask about access needs or self-identified neurotype where your study needs it.
• Preview first. Preview the interview yourself for free, by voice or text, and ideally ask a neurodivergent colleague or advisor to try it and point out anything confusing.

Koji does not replace co-design or in-person sessions where you need to observe someone in their own environment, and it does not provide its own text-size or contrast controls, so check the experience on the devices and settings your participants use. What it does is make the low-pressure, written-or-spoken, in-your-own-time format the default rather than an exception participants have to ask for.

FAQ

What does neurodivergent mean in user research? It describes people whose brains work differently from what is considered typical, including autistic people and people with ADHD, dyslexia, dyspraxia and related conditions. In research, it matters mainly because standard session formats are designed around neurotypical communication.

Should I ask participants whether they are neurodivergent? Only if your research question requires it. Asking everyone about access needs is usually better: it is less intrusive, it includes people who are undiagnosed, and it tells you what to change. If you do ask about neurotype, make it optional and treat it as sensitive data.

Are asynchronous interviews better for neurodivergent participants? Often, but not always. Asynchronous, self-paced interviews remove scheduling, cameras and time pressure, which many people find easier. Others prefer a live conversation. The most inclusive approach is to offer a choice.

How should I write interview questions for autistic participants? Be concrete and literal, ask one thing at a time, avoid idioms and hypotheticals, and ask about specific recent experiences. Sharing questions in advance helps many participants prepare.

Why do some interviews with neurodivergent participants feel like they went badly? Often because of a mismatch in communication styles on both sides, which Damian Milton called the double empathy problem. Crompton and colleagues (2020) found that information was lost faster in mixed autistic and non-autistic groups than in groups of either neurotype alone. Adjust the format and your questions before judging the participant.

Related Resources

• Structured Questions Guide
• Accessibility Research: Including Users with Disabilities
• Voice vs Text Interviews
• User Research with Older Adults
• How to Moderate User Interviews
• Research Ethics Guide

Sources

• Centers for Disease Control and Prevention (2024). ADHD diagnosis, treatment and telehealth use in adults, United States, October–November 2023. Morbidity and Mortality Weekly Report, 73(40).
• Shaw, K. A., et al. (2025). Prevalence and Early Identification of Autism Spectrum Disorder Among Children Aged 4 and 8 Years — ADDM Network, 16 Sites, United States, 2022. MMWR Surveillance Summaries, 74(SS-2). CDC.
• Milton, D. E. M. (2012). On the ontological status of autism: the "double empathy problem". Disability & Society, 27(6), 883–887.
• Crompton, C. J., Ropar, D., Evans-Williams, C. V. M., Flynn, E. G., & Fletcher-Watson, S. (2020). Autistic peer-to-peer information transfer is highly effective. Autism, 24(7), 1704–1712.